In the previous post, I've already defined the voicing rule as occurring in unstressed sequences of word-final *-VCᵊ#. However, I've also been pondering today on the possibility that the voicing here also involves laryngealization (aka creaky voicing)[1]. So now I will run some examples by you all to explain how this might explain not only word-final voicing in Late IE, but also the reflex of the thematic vowel, Szemerenyi's Law and maybe even the development of Narten presents. I know, I know, that's a lot on my plate but hear me out with a few examples showing the diachronic development:
Examples where voicing/laryngealization appears to occur:
*kahʷánasa 'dog (nom.)'Examples where voicing/laryngealization does not appear to occur:
*[kə'hʷanəsə]
*[kᵊ'wanᵊsᵊ] (Reduction)
*[kᵊ'wanᵊz̰ᵊ] (Laryngealization)
*[kwa:nz̰] (Syncope)
*kwānz (> PIE *ḱwōn)
*ʔékwasa 'horse (nom.)'
*['ʔekwəsə]
*['ʔekwəsᵊ] (Reduction)
*['ʔekwəz̰ᵊ] (Laryngealization)
*['ʔekwəz̰] (Syncope)
*ʔékwəz (> PIE *h₁éḱwos)
*bérata '(s)he carried'
*['berətə]
*['berᵊtᵊ] (Reduction)
*['berᵊd̰ᵊ] (Laryngealization)
*[be:rt] (Syncope + Analogical Levelling with presentive 3ps *bḗrti)
*bērt (> PIE *bʰērt)
*nákʷtasa 'night (nom.)'
*['nakʷtəsə]
*['nakʷtᵊsᵊ] (Reduction)
*[nakʷtᵊz̰ᵊ] (Laryngealization)
*[nakʷts] (Syncope; Lengthening blocked by consonant cluster; Assimilation)
*nakʷts (> PIE *nokʷts)
*kʷaisa/*kʷaita 'who?/what?'
*[kʷəisə]/*[kʷəitə]
*[kʷisᵊ]/*[kʷitᵊ] (Reduction)
*[kʷiz̰ᵊ]/*[kʷid̰ᵊ] (Laryngealization)
*[kʷiz̰]/*[kʷid̰] (Syncope)
*kʷiz/*kʷid̰ (> PIE *kʷis/*kʷid)
*pad̰sa 'foot (nom.)'
*['pad̰sə]
*['pad̰sᵊ] (Reduction)
*[pa:d̰z] (Syncope + Assimilation)
*pād̰z (> PIE *pōds)
*napátasa 'grandson (nom.)'
*[nə'patəsə]
*[nə'patᵊsᵊ] (Reduction + Resistance)
*[ne'pa:ts] (Syncope)
*nepā́ts (> PIE *nepōts)
*bératai '(s)he carries'
*['berətəi]
*['berᵊti] (Reduction)
*['be:rti] (Syncope)
*bḗrti (> PIE *bʰḗrti)
In the last example of the examples showing voicing/laryngealization, I show an instance of voicing in a completely unstressed word. The zerograde vowel *i in PIE *kʷid demonstrates that the word was unstressed during Syncope and this works well for the purposes of my new rule since apparently the preceding vowel is not as important as the vowel that follows it, which must be a supershort schwa. I presume that the difference in length between word-final supershort schwa and regular schwa is what triggered this change in the first place. The example of *h₁éḱwos 'horse' is meant to show the evolution of a thematic stem in the nominative. Its thematic vowel naturally then surfaces as *-o- because the nominative singular became voiced since the stage of Syncope. You may be skeptical of a voiced *-z contrasting with *-s in early Late IE (as I've been for some time) however it explains why PIE *-s is dropped in nominative singular *ḱwōn 'dog' while the *-s remains in the accusative plural ending *-ns. This can be solved by positing earlier *kwānz (from MIE *kahʷánasa with word-medial *-s-) and *-ns (from MIE *-am-as with word-final *-s) respectively. You may also note that certain vowels lengthen during Syncope in my above examples. I believe this is caused by the reduction of two supershort schwas in adjacent open syllables (i.e. *-V́C₁ᵊC₂ᵊ -> *-V̄́C₁C₂) via compensatory lengthening. This might then help to explain the rise of Narten presents which Jay Jasanoff considers to be most ancient. Personally, I don't believe that long vowels existed in the Mid IE period at all.
So whaddya say, people?! Isn't that dandy??? Sorry, it's getting me excited but all this is very complicated and making my head spin. I think this can work but we'll see what my fellow commenters say.
NOTES
[1] Compare with RESTLE, D., ZAEFFERER, D., & VENNEMANN, T. (2002). Sounds and systems: Studies in structure and change: A festschrift for Theo Vennemann. Trends in linguistics, 141. Berlin, Mouton de Gruyter. p.24 (see link): "Since only word-internal tense plosives were preglottalized, the Old Danish apocope produced such pairs as WJut. kat - kat' 'cat - cats' [kat - kaʔt] corresponding to St.Da. kat - katte." It's not quite the same but still eerily similar to the PIE examples.
UPDATES
(Jun 16 2008) I found a sweet Danish parallel involving similar laryngealization just a few hours after I posted this entry which I've added as a footnote. Enjoy!
Since your last post some of this has been spinning through my head.
ReplyDeleteAnd that the o-stems come from a schwa which for 'some' reason was different from the syncopation schwa's was clear too.
I can't help but feel that PIE looks so Proto-Germanic now :D
Only thin I'm curious about, is how do you plan to explain the difference between the appearance of a 'schwa' and a 'supershortschwa'.
After all the phonetic environments of
kə'hʷanəsə
and
'ʔekwəsə
These phonetic environments are so similar, it's hard to figure out why one retains a 'long schwa' in the penultimate syllable, and the other a 'short schwa' in an almost identical environment.
Phoenix: "These phonetic environments are so similar, it's hard to figure out why one retains a 'long schwa' in the penultimate syllable, and the other a 'short schwa' in an almost identical environment."
ReplyDeleteYes, but they're dissimilar structurally nonetheless since their respective stems are kə'hʷánə- and 'ʔékwə-. In the latter stem, the schwa resists deletion to avoid an illegal coda in resultant nominative **ʔekw-z that would otherwise develop.
Maybe I should add: We shouldn't expect Mid IE unstressed *-wa- to reduce to Late IE *u because this would require added metathesis (i.e. metathesis of schwa and *w). Only *au and *awa become Late IE *u.
ReplyDeleteI'd still like to understand why this alleged voicing occurred only in *-VCᵊ#. There does not seem to be any conditioning environment discernible that would motivate such a severe limitation.
ReplyDeleteRob: "I'd still like to understand why this alleged voicing occurred only in *-VCᵊ#. There does not seem to be any conditioning environment discernible that would motivate such a severe limitation."
ReplyDeleteYes, I'm aware that my idea is still rough and needs to be sufficiently ironed out and explained. The onus is on me.
However, I think that the example of Old Danish that I added in my footnotes may help in giving us helpful details as to what could have happened to create this word-final voicing pattern without it being "word-final".
I was just thinking yesterday that perhaps gemination occurred first in unstressed sequences of *-VCᵊ#. The gemination would require a preceding unstressed vowel and it would also require that the vowel after the consonant be a supershort schwa. Why? Let me explain.
The final supershort schwa, unlike other vowels, would have been growing shorter in duration over time and so gemination would in effect be the transfer of duration to the preceding consonant. This sort of gemination would also, in a sense, be a "splitting" of a consonant between syllables such that it is both the coda of the preceding syllable and the onset of the next (i.e. VCᵊ -> VCCᵊ).
Intervening gemination then would give us a reason why this only occured before word-final supershort schwa and didn't occur before other word-final vowels like *-i or *-ə (non-supershort schwa) which coincidentally weren't being shortened to oblivion and therefore could not have triggered this gemination.
So then, the following scenario might have occurred: *[-ətᵊ] -> *[-əttᵊ]. After Syncope, we would arrive at *[-ət:] which would trigger laryngealization like in the Danish example. Hence, *[-əd̰]. Consequently, *-əti just before Syncope would remain as is after Syncope because the *i is not being shortened and therefore no transfer of duration to the preceding consonant. Without gemination, no voicing!
Does this explanation now satisfy the people? :-)
Glen: "Does this explanation now satisfy the people? :-)"
ReplyDeleteSorry, but no.
For one thing, this gemination idea seems to me like multiplication of hypotheses. Second, I now realize that I should question your theory of Pre-IE syncope. To me it seems needlessly limited -- word-final vowels were hardly the only vowels which must have been elided in IE or Pre-IE. Furthermore, I still see no reason why the word-final position is so special as to warrant such a multiplication of hypotheses.
From where I am right now, a simpler point of view is this: IE underwent a vowel-reducing stage, implying strong stress timing, followed by a vowel-alternating stage, implying syllable timing. Not only that, but I posit that the latest stage(s) of IE did not even have syllable prominence on the word level; rather, it/they only had it on the phrase level.
Methinks you disagree, Rob, because you misunderstand my position. Let's discuss.
ReplyDeleteRob: "For one thing, this gemination idea seems to me like multiplication of hypotheses."
What hypotheses specifically are being multiplied?
Rob: "Second, I now realize that I should question your theory of Pre-IE syncope. To me it seems needlessly limited -- word-final vowels were hardly the only vowels which must have been elided in IE or Pre-IE."
But this is not my position. My view is that unstressed schwa in all positions was first reduced to supershort schwa in Mid IE (i.e. Reduction). However, *some* vowels resisted reduction and remained "full schwa" rather than supershort. It was the *supershort* schwas, again in *all* positions of a word, which were deleted during the later event of Syncope. So it seems that we are actually in agreement on that, putting aside these added details. (I know I should really make a pdf explaining all of these intricacies because it's even starting to confuse me, hehe.)
Rob: "Furthermore, I still see no reason why the word-final position is so special as to warrant such a multiplication of hypotheses."
I'm still not certain which hypotheses you're referring to per se. However, I see no way to avoid the fact that, in some way, MIE postclitics *-sa and *-ta ended up voiced to *-z and *-d in early Late IE once Syncope was said and done. At least, the latter suffix is clearly voiced but the former suffix (which becomes nominative *-s) is shown to be once voiced [-z] by the thematic vowel *-o- rather than *-e- in the thematic nominative. These facts prove conclusively to me that both these suffixes came to be voiced at the same time.
So if they came to be voiced, how was this done? What is the most efficient and credible means? Simply claiming that word-final *-t and *-s were voiced after Syncope, as I had previously theorized, doesn't cut it because such a phenomenon is rare and doesn't explain why *-s disappears in nominative *ḱwōn but not in the accusative plural ending *-ns.
This is why I believe that gemination occurring before Syncope as a means to transfer vocalic duration to the preceding consonant (as per my comment directly above) is a comparatively more credible hypothesis. Geminated consonants can and do become laryngealized (aka "are given creaky vocalization"), by the way.
Rob: "From where I am right now, a simpler point of view is this: IE underwent a vowel-reducing stage, implying strong stress timing, followed by a vowel-alternating stage, implying syllable timing."
Yes, this is basically what I believe as well. In fact, many (or most?) IEists agree that zerograding implies a stress accent in Pre-IE to account for its development, and then naturally, this stress accent would later become tonal by the time of PIE to account for the tonal accents found in Greek and Sanskrit.
Sorry it took me so long to write back, Glen.
ReplyDeleteGlen: "What hypotheses specifically are being multiplied?"
Off the top of my head: A-Epenthesis, Medial Voicing (When Later in Final Position), and now Medial Gemination on top of Medial Voicing.
Glen: "But this is not my position. My view is that unstressed schwa in all positions was first reduced to supershort schwa in Mid IE (i.e. Reduction)."
Okay, fair enough, except I'm not sure why you make a distinction between "supershort schwa" and phonetic zero.
Glen: "However, *some* vowels resisted reduction and remained "full schwa" rather than supershort."
Okay, what do you think were the phonological conditions for "full schwa" vs. "supershort schwa"?
Glen: "It was the *supershort* schwas, again in *all* positions of a word, which were deleted during the later event of Syncope. So it seems that we are actually in agreement on that, putting aside these added details. (I know I should really make a pdf explaining all of these intricacies because it's even starting to confuse me, hehe.)"
I think we seem fairly close to agreement. However, the devil is in the details. The question is under what conditions (if any!) could unstressed vowels be elided.
Glen: "I'm still not certain which hypotheses you're referring to per se. However, I see no way to avoid the fact that, in some way, MIE postclitics *-sa and *-ta ended up voiced to *-z and *-d in early Late IE once Syncope was said and done. At least, the latter suffix is clearly voiced but the former suffix (which becomes nominative *-s) is shown to be once voiced [-z] by the thematic vowel *-o- rather than *-e- in the thematic nominative. These facts prove conclusively to me that both these suffixes came to be voiced at the same time."
Of course, that all begs the question as to whether the athematic nom. sg. ending *-s and athematic pronominal nom./acc. (better "abs." for absolutive) sg. ending *-d are indeed from postclitic demonstratives/articles, as opposed to coming from some other source(s).
Rob: "Sorry it took me so long to write back, Glen."
ReplyDeleteNo problem. Time is no factor. Let the Games continue!
Rob: "Off the top of my head: A-Epenthesis, Medial Voicing (When Later in Final Position), and now Medial Gemination on top of Medial Voicing."
You need to first show how these hypotheses are insufficient before your objection is valid. Simply listing my hypotheses as if the very idea of a theory consisting of a number of interrelated hypotheses were somehow objectionable a priori is an illogical conclusion.
Since a-Epenthesis, for example, is based on Jens Rasmussen's published contributions (i.e. his *O-infix hypothesis), you need to first confront the data he uses to support it (and which is in part supporting my reinterpretation of his rule).
Medial voicing is also based on Rasmussen's clever observations on the underlying pattern of ablaut of the thematic vowel: *e before unvoiced segments and *o before voiced ones. So if the only exception is that *o precedes nominative singular *-s, it's likelier that this ending was once voiced like all other examples. And then, if animate nominative *-s < *-z and inanimate nomino-acc. *-d are most optimally explained etymologically and semantically by animate *so and inanimate *tod, we are forced to conclude that something has triggered voicing of expected word-final *-s and *-t. Yet, since word-final voicing is quite rare except in word-medial, intervocalic positions (see Blevins, Evolutionary Phonology: The Emergence of Sound Patterns (2004), page 109), this voicing can only most rationally precede Syncope and therefore dates to the Mid IE period. The trigger of voicing is therefore most likely to be word-medial, intervocalic laryngealization of geminated *-ss- and *-tt- before there was loss of the final vowel. I find no chain of reasoning more efficient so far that doesn't violate linguistic norms and tendencies. Do you?
Again, where is a thorough rebuttal which is required of you to convincingly and completely disprove the necessity of the hypotheses that I suggest? Now it's time to do your homework too.
Rob: "[...] except I'm not sure why you make a distinction between 'supershort schwa' and phonetic zero."
I suggest now that there's first a period of Reduction because it explains more precisely the details behind a-Epenthesis. It's a matter of increasing precision in my evolving theory and psychotic pursuit for ultimate intellectual perfection.
You see, it seems that a-Epenthesis is always triggered when an awkward word-initial cluster would otherwise develop. However, if we say only that the unstressed schwa simply disappeared (i.e. straight from schwas to Syncope), then there's never a point at which the word-initial cluster would threaten to develop in order to trigger the rule in the first place! So instead, we need to acknowledge a brief intermediary period of Reduction that converts *most* schwas to supershort schwa while preserving a full schwa in a few cases.
Naturally, once Reduction would have happened, the schwas became short enough to trigger a-Epenthesis because the difference between true CCC- and CCC- seperated with internal non-moraic supershort schwas is acoustically minute (but an important difference nonetheless for the changing syllable structure of late Mid IE). In other words, at most, the awkward word-initial clusters that were to be avoided would have only developed in a phonemic sense, yet never on the phonetic level. Before that could ever happen, the inserted instances of *a came to the rescue to break things up.
Rob: "Okay, what do you think were the phonological conditions for 'full schwa' vs. 'supershort schwa'?"
The reduction of vocalic length to supershort schwa was the default by far and preservation is the more minor exception to the rule. The exception is triggered by more than one factor. One exception is Suffix Resistance whereby a monosyllabic suffix of the form CV avoids reduction in a word, thereby preserving its vowel and consequently its syllabicity. This particular exception may simply be part of a general rule that avoids unstressed *-C.CV- being reduced to *-C.C- (n.b. the period marks syllable boundaries). Another exception is Paradigmatic Resistance which prevents roots from becoming completely asyllabic in paradigms. I keep using the alternation of accusative *pódm 'foot' and genitive *pedós as a clear proof of this avoidance of asyllabicity.
Rob: "Of course, that all begs the question as to whether the athematic nom. sg. ending *-s and athematic pronominal nom./acc. (better "abs." for absolutive) sg. ending *-d are indeed from postclitic demonstratives/articles, as opposed to coming from some other source(s)."
This in turn begs the question why you don't specify what these other sources are and specify constructively how they are more efficient than the solutions I provide here. Without doing this, all ideas ever created can be criticized with these imaginary and vague alternatives... and that would be the same nihilist tactics as seen on FoxNews. How dreadful! :-)
On a new counterargument, are you sure you can get this layer of gemination completely out of the way before those *tt that > *st / *ss kick in? IIUC these will also be created by your Syncope, if not before that.
ReplyDeleteI'd suggest direct insertion of a glottal stop, rather than gemination, except that's not really a process associated with an unstressed position. But neither is gemination - your comparision points from Danish involve a preceding stressed vowel. And I still say plain voicing could work as well (see my latest reply on the Szeremenyi's Law post).
There are examples of laryngealization not tied to stress (eg. English) but I've only ever seen this involving stops - and would not retain a *-t / *-d distinction since a formerly medial position for the latter is nowhere to be seen anymore.
For *-s, there's however the possibility that *z is the original form here and *so is simply also devoiced.
And for your latest debate here: wouldn't Suffix Resistance predict these suffixes not reducing to single consonants?
Tropylium: "[...] are you sure you can get this layer of gemination completely out of the way before those *tt that > *st / *ss kick in?"
ReplyDeleteAh, are you by chance referring to the sibilantization between *-TT- in the pronunciation of PIE words like *sédtos 'seat'? I figure that this sibilantization need not occur during Syncope necessarily.
Tropylium: "I'd suggest direct insertion of a glottal stop,[...]"
Effectively, you'd be adding ejectives to a phonology that already has plain-voiced and creaky-voiced stops. This seems relatively less economic as a solution.
Tropylium: "But neither is gemination - your comparision points from Danish involve a preceding stressed vowel."
Can you explain what linguistic processes require that stress be a factor in gemination or laryngealization? Please also note that the syllable in question would have once been given secondary stress.
Tropylium: "And I still say plain voicing could work as well (see my latest reply on the Szeremenyi's Law post)."
Yet how do you explain *z and creaky-voiced *d as outcomes then without having to theorize extra assumptions such as the neutralization of word-final stops that apparently did not occur based on the PIE data? You haven't explained how your idea concerns verb stems ending in both creaky-voiced and plain-voiced stops in PIE's 2ps endingless imperative.
Tropylium: "For *-s, there's however the possibility that *z is the original form here and *so is simply also devoiced."
Not if Uralic 3ps *sa is cognate with Mid IE *sa (becoming PIE *so via Vowel Shift).
Tropylium: "And for your latest debate here: wouldn't Suffix Resistance predict these suffixes not reducing to single consonants?"
Yes, it would. So the reduction of these syllables is an exception to an exception. The proof of their rebellion against Suffix Resistance is the added length of the previous syllable (as found in the nominative singular and probably also in Narten presents), a remnant of the vowel that wasn't supposed to be deleted. This is explained away quite simply as garden-variety compensatory lengthening, a very common process at work in languages worldwide.
Concerning gemination and its purported association to stress, consider Old English wēstenne 'desert (dat.sg.)' < wēstenje despite wīte 'punishment' < wītje.
ReplyDeleteEffectively, you'd be adding ejectives to a phonology that already has plain-voiced and creaky-voiced stops. This seems relatively less economic as a solution.
ReplyDeleteI don't follo. You suggested geminate > glottalized > creaky. I propose that the geminate step is unnecessary.
Can you explain what linguistic processes require that stress be a factor in gemination or laryngealization? Please also note that the syllable in question would have once been given secondary stress.
No, I'm not saying that stress is required, just that you can't require the *lack* of stress to trigger gemination. Secondary stress would be fine in principle, but it clashes with the exceptional reduction...
You haven't explained how your idea concerns verb stems ending in both creaky-voiced and plain-voiced stops in PIE's 2ps endingless imperative.
Ack, shit, I was only looking at suffixes.
I didn't manage to quickly locate any actual examples of this specific distinction, but yes, this would need looking into. The possibility of paradigmatic reversion would also come into play at this point, tho.
Not if Uralic 3ps *sa is cognate with Mid IE *sa
Proto-Uralic has no voice distinction whatsoever and can thus hardly be used to debunk any specific phonation.
And BTW, are you saying there's basically two distinct layers of syncope, one of which causes compensatory lengthening and one of which doesn't?
Tropylium: "You suggested geminate > glottalized > creaky."
ReplyDeleteNo, I only suggested geminate → creaky. Both gemination and creaky phonation are a form of fortition. Why do you feel that glottalization is necessary? At this stage in Mid IE, ejectives would have already lenited to creaky stops. So this is why I say that you're adding extra phonemes to the inventory and that this is uneconomical.
Tropylium: "No, I'm not saying that stress is required, just that you can't require the *lack* of stress to trigger gemination."
I never said such a thing. I'm open to the possibility however that this may be triggered by a former secondary stress but then, as we both can see, this means that I must push gemination further back in time to before Reduction. An interesting possibility. I'll have to ponder this today.
Tropylium: "Ack, shit, I was only looking at suffixes."
Hehe. Yes! There's a lot of pesky details to keep track of. I think that the reduction of phonetic possibilities in suffixes is natural in any language. So the fact that PIE suffixes lack things like *-p and *-dh, despite these sounds being allowed word-finally in some wordforms, isn't particularly odd in my view.
Tropylium: "Proto-Uralic has no voice distinction whatsoever and can thus hardly be used to debunk any specific phonation."
You found the logical loophole, I see. While this is true, Altaic has a voice distinction and yet many Nostraticists such as Illich-Svitych and Dogopolski see a cognate series involving a pronoun in *s-, not **z-. Your hypothesis therefore flies against the available data unless you can find a better cognate series that demonstrates *z-.
Tropylium: "And BTW, are you saying there's basically two distinct layers of syncope, one of which causes compensatory lengthening and one of which doesn't?"
Effectively, I suppose, yes. Lengthening is only triggered by monosyllabification (an exceedingly common source of vowel lengthening in other languages such as Latin) and Clipping (which is afterall an exception to an exception of Syncope, defying the Suffix Resistance rule). I've therefore explained precisely in what cases lengthening is to be expected.